Radia Perlman: Stabilizing Networks through a Winding Career

November 30, 2017
Radia Perlman

Radia Perlman doesn’t like computers. But she’s always been a problem-solver. And computers and software proved to have plenty of problems to hold her interest.

It was only through a series of unexpected and unsolicited opportunities, however, that she even got into the field of computer programming, where she went on to earn renown for solving some of the most vexing problems of early networking and develop a groundbreaking thesis that is still the staple of college curriculum today.

Perlman’s initial aversion to computer programming came in high school, when a teacher signed a few students up for a class at Stevens Institute of Technology.

“After having never had any problem with any class in school (with the possible exception of phys ed), I walked into that class and people were talking about how they built ham radios when they were seven,” she said. “I had no idea what a ham radio was. I certainly couldn’t build one.”

“Then, when they asked questions, they were using fancy words like ‘input.’ I had no idea what that was. I just decided I was so far behind, I would never catch up. And I got nothing out of that class.”

So, Perlman said, she decided to stick with studies involving books, papers and pencils. That is, until she was an undergrad at the Massachusetts Institute of Technology, when a teaching assistant in her physics class asked her to be his programmer.

“I said, ‘I don’t know how to program,’” Perlman recalled. “He said, ‘Yes, I know that. That’s why I’m asking you, because I don’t have any money to pay you and if you knew how to program you would expect to be paid.’ He said he was sure I would learn.”

She not only learned, she said, she enjoyed it. So, she got a part-time job in MIT’s artificial intelligence lab as a systems programmer.

Her introduction to network protocols was just by chance.  As a graduate student in math, she’d completed all her requirements other than a thesis.

“I had no idea how to write a thesis,” Perlman said. “… I couldn’t imagine doing research. I knocked on few doors and said I needed an advisor. They all basically said that they were important busy people.”

Then she confided in an old friend that she was not happy in grad school, because she had no idea how to find an advisor and get started on a thesis.  Her friend said, “join our group at BBN.”  That involved designing protocols for the packet radio network.

“I realized I really loved it,” she said. “What is really cool about network protocols is having a lot of independent little things all doing a piece of a program, and it all fits together. People think in different ways. I happen to be able to envision lots of things happening in parallel.”

From there, Perlman said, she got her dream job at Digital Equipment Corporation (Digital) where she said her first contribution to the Internet was introducing the notion of network self-stabilization. After recognizing a weakness in ARPAnet Protocols that would enable one bad actor to permanently disable the network, she raised alarms and went to develop the protocols that enabled the ARPAnet and today’s Internet to recover by itself once the bad guy is disconnected. This innovation was one of the important contributions Perlman made to what is called “link state routing protocols”, which are used for routing IP today. The specific protocol she designed for DECnet was standardized and renamed IS-IS, which continues to be widely deployed today.  Although IS-IS can route many types of protocols, the concepts were copied into an IP-only version known as OSPF.

While Perlman thinks her contributions that make routing protocols resilient, scalable, and manageable are her most important contributions, she is perhaps best known for designing the spanning tree algorithm in the early 1980’s, which transformed Ethernet from the original limited-scalability, single-wire CSMA/CD, into a protocol that can handle large clouds.  Later, she improved upon the spanning tree by designing TRILL (Transparent Interconnection of Lots of Links), which allowed the Ethernet to make optimal use of bandwidth.

She did wind up returning to graduate school and getting her PhD, due to, as she puts it, “the strangest reason anyone has had for going back to grad school.” Although she was thriving professionally at Digital, and successfully juggling young children and career, as she said, “I always had the world’s worst cold.  Wherever I went, I had to carry a wastepaper basket and a big box of tissues.  Eventually I went to a doctor and asked if he could give me anything to help.  He asked how long I’d had the cold, and I said, “3 or 4 years.” He pointed out that was not a cold, but allergies.  “I eventually figured out it was due to cigarette smoke.  Back then, people smoked in the building.  I’d never been exposed to cigarette smoke before Digital.  People told me to be patient, because Digital was working on the issue. I decided to take advantage of the company’s continuing education program, which covered all expenses and full salary for two years, hiding out in grad school while they solved the smoking issue.”

She ended up writing a ground-breaking doctoral thesis on how to make networks resilient even if malicious trusted components are actively trying to sabotage the network. This thesis is still widely used in college curriculum today.

After grad school, Perlman returned to Digital, and also wrote the popular textbook Interconnections. “Writing the book had an amazing effect on my career. I’m not very good about being self-promoting, but given that the book became the book that everyone learned the field from, I didn’t need to act impressive for people to take me seriously.”

After writing Interconnections, she co-wrote the book Network Security with her colleague Charlie Kaufman. That book was also very successful, and as Perlman said, “we wrote a book together and didn’t kill each other, so we are soulmates.”  She is currently at Dell EMC, living in the Seattle area.

Add new comment

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
This question is for testing whether you are a human visitor and to prevent automated spam submissions.